Wendt 1987

International Organization
http://journals.cambridge.org/INO
Additional services for International

Organization:
Email alerts: Click here

Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here
The agent-structure problem in international

relations theory
Alexander E. Wendt
International Organization / Volume 41 / Issue 03 / June 1987, pp 335 - 370

DOI: 10.1017/S002081830002751X, Published online: 22 May 2009
Link to this article: http://journals.cambridge.org/abstract_S002081830002751X
How to cite this article:

Alexander E. Wendt (1987). The agent-structure problem in international relations
theory. International Organization, 41, pp 335-370 doi:10.1017/
S002081830002751X
Request Permissions : Click here
Downloaded from http://journals.cambridge.org/INO, IP address: 128.184.220.23 on 13 Mar 2015

The agent-structure problem
in international relations theory
Alexander E. Wendt
Two theories, neorealism and world-system theory, strongly influence con-

temporary academic discourse about international relations. Both claim to
provide "structural" explanations of how states behave in the international
system. Despite their common commitment to structural analysis, however,
their understanding of system "structure," and therefore of structural expla-
nation, is quite different. Neorealists define international system structures
in terms of the observable attributes of their member states (the "distribu-
tion of capabilities"), and as a result, they understand the explanatory role
of those structures in individualist terms as constraining the choices of pre-
existing state actors. World-system theorists, on the other hand, define inter-
national system structures in terms of the fundamental organizing principles
of the capitalist world economy which underlie and constitute states, and
thus they understand the explanatory role of structures in structuralist terms
as generating state actors themselves. These differences, and their implica-
tions, have yet to be explicated in the international relations literature.1 In
An earlier version of this article was presented at the 1986 meeting of the International
Studies Association. I want to thank Hay ward Alker, Richard Ashley, Raymond Duvall, Jeffrey
Isaac, Brian Job, Stephen Krasner, Peter Manicas, David Sylvan, Jutta Weldes, and two
anonymous referees for their helpful comments and suggestions on previous drafts.
1. There are a number of discussions of the meanings and uses of "structural theory" in
neorealism and world-system theory, but as far as I know, none explicitly compares or differ-
entiates the neorealist and world-system approaches to structure and structural analysis. On
neorealism see, for example, Kenneth Waltz, Theory of International Politics (Reading, Mass.:
Addison-Wesley, 1979), and Robert Keohane, "Theory of World Politics: Structural Realism
and Beyond," in Ada Finifter, ed., Political Science: The State of the Discipline (Washington,
D.C.: APSA, 1983). The best critique of neorealism's conception of structure is Richard Ash-
ley's "The Poverty of Neorealism," International Organization 38 (Spring 1984), pp. 225-86.
On world-system theory see Immanuel Wallerstein, "The Rise and Future Demise of the World
Capitalist System: Concepts for Comparative Analysis," Comparative Studies in Society and
History 16 (September 1974), pp. 387-415, and Christopher Chase-Dunn and Richard Rubinson,
"Toward a Structural Perspective on the World-System," Politics and Society 1 (no. 4, 1977),
pp. 453-76. The critique of world-system theory that comes closest to my concerns in this
article is probably Theda Skocpol's "Wallerstein's World Capitalist System: A Theoretical and
Historical Critique," American Journal of Sociology 82 (March 1977), pp. 1075-90.
International Organization 41, 3, Summer 1987

© 1987 by the World Peace Foundation and the Massachusetts Institute of Technology
336 International Organization
this article, I want to begin to clarify and contrast the nature of structural
analysis in each of these two traditions. My primary interest, however, is to
critique the conceptions of structural theory found in each of them, and to
use this critique to motivate the development of a new approach to structural
theorizing about international relations adapted from the work of "structura-
tion theorists" in sociology.2 This approach and the research agenda it im-
plies, in turn, require a foundation in realist philosophy of science (or
"scientific realism"3), arguably the "new orthodoxy" in the philosophy of
natural science, but as yet largely unacknowledged by political scientists.4
As structural theories of international relations, neorealism and world-
system theory differ, and thus might be compared, along a number of dimen-
sions: substantive claims, predictive power, scope, and parsimony, among
others. While these differences are important, they are, I think, strongly
conditioned by a more fundamental difference of ontology: neorealism em-
bodies an individualist ontology, while world-system theory embodies a
holistic one. A useful way to capture the nature and implications of this
difference is to evaluate the two theories in terms of their underlying as-
sumptions about the relationship of system structures to human agents.
Despite their commitment to "structural" rather than "agentic" theorizing,
like all structural theories they both presuppose some theory of what is being
2. The term "structuration theory" is sometimes narrowly identified with the work of An-
thony Giddens, who has articulated its basic problematic in his Central Problems in Social
Theory (Berkeley: University of California Press, 1979) and The Constitution of Society: Out-
line of the Theory of Structuration (Cambridge, U.K.: Polity Press, 1984). In "On the Determi-
nation of Social Action in Space and Time," Society and Space 1 (March 1983), pp. 23-57,
however, Nigel Thrift uses the term more broadly as a generic label for a group of social
theories which share certain fundamental assumptions about the agent-structure relationship;
this group includes, but is not limited to, Pierre Bourdieu, Outline of a Theory of Practice
(Cambridge: Cambridge University Press, 1977), Roy Bhaskar, The Possibility of Naturalism
(Brighton, U.K.: Harvester Press, 1979), and Derek Layder, Structure, Interaction, and Social
Theory (London: Routledge & Kegan Paul, 1981). Since my purpose in this paper is less to
advance Giddens's ideas (indeed, I will rely more on Bhaskar than Giddens) than to demon-
strate the relevance of the overall problematic for international relations theory, I shall follow
Thrift's more inclusive use of the term.
3. Scientific realism (or simply "realism") is not related to political realism or neorealism in
international relations.
4. Whether or not scientific realism is the "new orthodoxy" in the philosophy of natural
science is undoubtedly a contentious issue among realists and empiricists, but it has in any case
made sufficient inroads that the Minnesota Center for the Philosophy of Science, long an
important bastion of empiricism, held a year-long institute in 1985/86 which, among other
things, focused explicitly on that question. American political scientists generally seem to be
unaware of or uninterested in this debate and its potential implications for political science. To
my knowledge, the only discussions of scientific realism in international relations are British:
John Maclean, "Marxist Epistemology, Explanations of Change and the Study of International
Relations," in Barry Buzan and R. J. Barry Jones, eds., Change in the Study of International
Relations: The Evaded Dimension (London: Frances Pinter, 1981), pp. 46-67, and Richard
Little, "The Systems Approach," in Steve Smith, ed., International Relations: British and
American Perspectives (Oxford: Blackwell, 1985), pp. 79-91.
Agent-structure problem 337
structured, human or organizational agents, and of their relationship to so-

cial structures. Put more generally, all social scientific theories embody an at
least implicit solution to the "agent-structure problem," which situates
agents and social structures in relation to one another. These solutions help
determine a theory's understanding of, and the relative explanatory impor-
tance it attaches to, structural analysis. While generating very different
understandings of structural theory, however, I shall argue that the neoreal-
ist and world-system solutions to the agent-structure problem are, in at least
one respect, very similar, and that this similarity creates a common funda-
mental weakness in these theories as "structural" approaches to interna-
tional relations. Structuration theory, in turn, is a response to this common
weakness which both subsumes and points beyond neorealism and world-
system theory.
In Section 1, I examine the nature of the agent-structure "problem" and
briefly identify the principal kinds of solutions to it. I argue in Section 2 that
neorealism and world-system theory embody two of these solutions, the
methodological individualist and structuralist ones, respectively. Despite
important differences between them, each of these approaches solves the
agent-structure problem by making either state agents or system structures
ontologically primitive units. The resulting effect on neorealism and world-
system theory is an inability to explain the properties and causal powers of
their primary units of analysis, a weakness which seriously undermines their
potential explanations of state action. This situation can be prevented by
adopting an approach to the agent-structure problem which does not pre-
clude a priori making both agents and structures "problematic" or "de-
pendent variables." In Section 3, I describe this third, structurationist
approach, and its foundations in realist philosophy of science. Since the
utility of structuration theory as a meta-theoretical framework for interna-
tional relations ultimately depends on its ability to enrich substantive
theorizing and concrete empirical research, its value cannot be convincingly
demonstrated in a programmatic article such as this one. It is possible,
however, to indicate some of the changes which a structurationist perspec-
tive suggests are necessary in the contemporary research agenda in interna-
tional relations. Towards this end, in Section 4, I examine some general
epistemological and theoretical implications of structuration theory for the
explanation of state action. In the conclusion, I return to some implications
of scientific realism for social scientific research.
1. The agent-structure problem
The agent-structure problem has its origins in two truisms about social life
which underlie most social scientific inquiry: 1) human beings and their
organizations are purposeful actors whose actions help reproduce or trans-

form the society in which they live; and 2) society is made up of social
relationships, which structure the interactions between these purposeful ac-
tors. Taken together these truisms suggest that human agents and social
structures are, in one way or another, theoretically interdependent or mutu-
ally implicating entities. Thus, the analysis of action invokes an at least
implicit understanding of particular social relationships (or "rules of the
game") in which the action is set—just as the analysis of social structures
invokes some understanding of the actors whose relationships make up the
structural context. It is then a plausible step to believe that the properties of
agents and those of social structures are both relevant to explanations of
social behavior. And in fact, although in very different ways, neorealism and
world-system theory do use the properties both of states (powers, interests)
and of system structures (polarity, relations of unequal exchange) to explain
state behavior. The "problem" with all this is that we lack a self-evident
way to conceptualize these entities and their relationship. This absence of a
single, immediately compelling conception of the agent-structure relation
has spawned a variety of conceptualizations of the relationship across the
social sciences,5 each reflecting the particular philosophical and practical
commitments of its parent theoretical discourse. (My own adoption of the
language of "agents" and "structures," therefore, is not theory-neutral.)
Despite their many differences, however, the "agent-structure," "parts-
whole," "actor-system," and "micro-macro" problems all reflect the same
meta-theoretical imperative—the need to adopt, for the purpose of ex-
plaining social behavior, some conceptualization of the ontological and ex-
5. The agent-structure problem has, in various guises, recently emerged as something of a

cottage industry throughout the social sciences. For a sampling of this work: in geography, see
Derek Gregory, "Human Agency and Human Geography," Transactions of the Institute of
British Geographers 6 (no. 1, 1981), pp. 1-18, and Derek Gregory and John Urry, eds., Social
Relations and Spatial Structures (London: MacMillan, 1985); in sociology, in addition to the
work of Giddens and Bhaskar already cited, see Alan Dawe, "Theories of Social Action," in
Tom Bottomore and Robert Nisbet, eds., A History of Sociological Analysis (London:
Heinemann, 1979), and Karin Knorr-Cetina and Aaron Cicourel, eds., Advances in Social
Theory: Toward an Integration of Micro and Macro-Sociologies (London: Routledge & Regan
Paul, 1981); in social history, see Philip Abrams, Historical Sociology (Ithaca: Cornell Univer-
sity Press, 1982), and Christopher Lloyd, Explanation in Social History (Oxford: Blackwell,
1986); in the philosophy of social science, see John O'Neill, ed., Modes of Individualism and
Collectivism (New York: St. Martins, 1973), and David-Hillel Ruben, The Metaphysics of the
Social World (London: Routledge & Kegan Paul, 1985); in Marxist theory, see Edward Thomp-
son's polemic in The Poverty of Theory and Other Essays (New York: Monthly Review Press,
1978) against the structural Marxism of Louis Althusser, and the commentaries on this debate
by Perry Anderson, Arguments within English Marxism (London: Verso, 1980), and Nicos
Mouzelis, "Reductionism in Marxist Theory," Telos 45 (Fall 1980), pp. 173-85; and in interna-
tional relations, see Waltz, Theory of International Politics, and James Rosenau, "Before
Cooperation: Hegemons, Regimes, and Habit-Driven Actors in World Politics," International
Organization 40 (Autumn 1986), pp. 849-94.
planatory relationship between social actors or agents (in this case, states)6
and societal structures (in this case, the international system).
The agent-structure problem is really two interrelated problems, one on-
tological and the other epistemological. The first, and more fundamental,
issue concerns the nature of both agents and structures and, because they
are in some way mutually implicating, of their interrelationship. In other
words, what kind of entities are these (or, in the case of social structures, are
they entities at all?), and how are they interrelated? There are two basic
ways to approach this question: by making one unit of analysis ontologically
primitive, or by giving them equal and therefore irreducible ontological
status. Depending on which entity is made primitive, these approaches gen-
erate three possible answers to the ontological question, which I will call
individualism, structuralism, and structurationism. Neorealism and world-
system theory embody, respectively, the first two of these positions, both of
which ultimately reduce one unit of analysis to the other. Thus, neorealists
reduce the structure of the state system to the properties and interactions of
its constituent elements, states, while world-system theorists reduce state
(and class) agents to effects of the reproduction requirements of the capi-
talist world system. The structurationist approach, on the other hand, tries
to avoid what I shall argue are the negative consequences of individualism
and structuralism by giving agents and structures equal ontological status.
Far from being a mindless synthesis of the "best of both worlds," however,
the structuration project requires a very particular conceptualization of the
agent-structure relationship. This conceptualization forces us to rethink the
fundamental properties of (state) agents and system structures. In turn, it
permits us to use agents and structures to explain some of the key properties
of each as effects of the other, to see agents and structures as "co-
determined" or "mutually constituted" entities.
The manner in which a social theory addresses these ontological issues
conditions its approach to the epistemological aspect of the agent-structure
problem, the choice and integration of different types of explanations within
theories of social behavior. This problem actually raises two epistemological
issues. The first is the choice of the form of explanation corresponding
respectively to agents and structures. This choice depends largely on the
kinds of properties of agents and structures that have been deemed causally
6. Recent theoretical work has conceptualized the state both as an agent and as a structure;
see, for example, Roger Benjamin and Raymond Duvall, "The Capitalist State in Context," in
Roger Benjamin and Stephen Elkin, eds., The Democratic State (Lawrence, Kans.: University
of Kansas Press, 1985), pp. 19-57. For purposes of this paper, 1 assume with neorealists that the
state is an agent, a move which can be justified in part because the organizing principles of the
state system constitute states as individual choice-making units which are responsible for their
actions. My subsequent arguments about the way in which system structures constitute states
as agents should not, however, be seen as excluding a conception of the state as a structure of
political authority in which governmental agents are in turn embedded.
significant. Thus, approaches to social inquiry that conceive of human be-

ings as reflective, goal-directed subjects, such as rational choice theory,
generate agent-explanations that are, broadly speaking, "interpretive"—
that is, cast in terms of the goals, beliefs, and self-understandings of agents.
On the other hand, approaches that conceive of human beings as nothing
more than complex organisms processing stimuli—such as behaviorism—
generate agent-explanations that are more mechanistically causal in form.
The situation is similar with respect to "structural" explanations. Social
theories that reduce system structures to the properties of individuals
usually construe the explanatory role of structures as one of constrain-
ing the choices of pre-existing agents, while those that conceptualize system
structures as irreducible entities underlying agents typically understand
structures as generating or explaining agents themselves. The second
epistemological issue concerns the relative importance of agent-expla-
nations and structure-explanations, of whatever type, in social theory. This
issue is of secondary importance in this article because neorealists and
world-system theorists agree that an adequate international relations theory
must be more structure- than agent-oriented. They understand this require-
ment in very different ways, however—a disagreement which I will later
show is linked to the way they approach the ontological dimension of the
agent-structure problem.
2. Reductionism and reification

in international relations theory
In this section I want to: 1) compare the conceptions of "structural" theory

found in neorealism and world-system theory; and 2) show that, despite
important differences, these conceptions share a common approach to the
agent-structure problem, and that this approach precludes an explanation of
the essential properties of their respective primitive units. This inability
leads to assumptions about primitive units that are without theoretical foun-
dation, a move which in turn undermines the theories' explanations of state
action in the international system. I shall argue that this common limitation
is a function of the basic assumptions and internal logic of each theory's
approach to the agent-structure problem, and that they therefore cannot
overcome it within the terms of their basic ontological and epistemological
commitments.
a. Neorealism
On the surface, at least, neorealists have strong structural and anti-
reductionist commitments. In his discussion of the nature of systemic and
reductionist theories, Kenneth Waltz defines the latter as theories which

explain the foreign policy behavior of states exclusively in terms of causes at
the national level of analysis.7 Lenin's theory of imperialism, for example, is
reductionist because it explains expansionist behavior in terms of the ac-
cumulation dynamics of national capitalism. Waltz criticizes such theories,
correctly I think, for ignoring the intervening role played by international
system structures in the translation of domestic imperatives into foreign
policy behavior. Neorealists avoid this type of reductionism by adopting the
systemic (but, I shall argue, not "structural") logic and conceptual ap-
paratus of micro-economic theory.8 This move permits neorealists to inte-
grate within a coherent theoretical framework the state-centric approach of
classical political realism with the systemic approach of international sys-
tems theory, and thus to develop a conception of the agent-structure rela-
tionship in international relations which recognizes the causal role of both
state agents and system structures.
In view of neorealists' desire to avoid micro-level reductionism, however,
it is ironic that their solution to the agent-structure problem is, in a different
and deeper sense, reductionist. The kind of "reductionism" which neoreal-
ists oppose is defined as theory which tries to explain behavior in terms of
strictly agent-level properties. This rejection of what might be called explan-
atory reductionism does not in itself, however, impose any particular restric-
tion on the ontological issue of how system structures should be defined,
since an opposition to agent-level explanations is analytically independent of
how system structures, once recognized as causally significant, should be
theorized. Thus, neorealists' individualist definition of the structure of the
international system as reducible to the properties of states9—to the distri-
bution of capabilities—is perfectly consistent with the important role that
7. Waltz, Theory of International Politics, p. 18.

8. Robert Gilpin, War and Change in World Politics (Princeton: Princeton University Press,
1981); Richard Ashley, "Three Modes of Economism," International Studies Quarterly 27
(December 1983), pp. 463-96; Robert Keohane, After Hegemony: Cooperation and Discord in
the World Political Economy (Princeton: Princeton University Press, 1984); Duncan Snidal,
"The Game Theory of International Politics," World Politics 38 (October 1985), pp. 25-57.
9. Ashley thoroughly critiques the individualist (and empiricist) foundations of the neorealist
conception of international system structure in his "Poverty of Neorealism," especially pp.
238-42. It is important to keep in mind, however, that in Theory of International Politics, Waltz
starts out with three defining features of political structures: 1) the principle according to which
they are organized, 2) the differentiation of units and their functions, and 3) the distribution of
capabilities across units. This definition can be used to support a generative approach to struc-
tural theorizing, as John Ruggie shows in his Durkheimian reconstruction of Waltz in "Con-
tinuity and Transformation in the World Polity: Toward a Neorealist Synthesis," World Politics
35 (January 1983), pp. 261-85. Despite this promising beginning, however, Waltz and other
neorealists argue that the first two features of this definition don't apply to international political
structures, leaving us in practice with an individualist conception of structure as the distribution
of capabilities. For an argument that links this result to a lingering neorealist commitment to
positivism, see Little, "The Systems Approach."
system structures play in neorealist explanations of state behavior. Indeed,

in both its decision- and game-theoretic versions10 neorealism, like micro-
economics, is characterized by "situational determinism," by a model of
action in which rational behavior is conditioned or even determined by the
structure of choice situations.11 The point is not that neorealists engage in
explanatory reductionism (which they do not), but rather that their definition
of system structure is characterized by ontological reductionism. This defini-
tion leads to an understanding of system structures as only constraining the
agency of preexisting states, rather than, as in world-system and structura-
tion theory, as generating state agents themselves. This follows inevitably
from its (ontologically) reductionist definition of system structure: system
structures cannot generate agents if they are reduced to the properties of
agents in the first place. Despite its strongly systemic focus, then, neoreal-
ism's view of the explanatory role of system structures is decidedly state- or
agent-centric. It sees system structures in the manner in which they appear
to states—as given, external constraints on their actions—rather than as
conditions of possibility for state action.
From the perspective of international relations theory, the most important
weakness of neorealism's individualist approach to the agent-structure prob-
lem is that it fails to provide a basis for developing an explicit theory of the
state. It is not hard to see why this must be the case. Theorizing about a
particular kind of individual unit, like the state, can take either of two forms
(or both). The first, "reductionist" strategy is to explain the individual in
terms of its internal organizational structure.12 While this approach may
explain some of the causal properties of the individual, it neglects the ir-
reducibly social content of many individual-level predicates.13 Thus, while
the internal physiological structure of a capitalist may explain some disposi-
tions and actions, we cannot explain his or her behavior as that of a "capi-
10. Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press, 1981);
Snidal, "The Game Theory of International Politics." Despite important differences between
the two versions over the conceptualization of choice situations in international relations, both
are based on an individualist definition of the structure of the international system as the
distribution of capabilities.
11. See Spiro Latsis, "Situational Determinism in Economics," British Journal for the Phi-
losophy of Science 23 (August 1972), pp. 207-45, and the reply by Fritz Machlup, "Situational
Determinism in Economics," British Journal for the Philosophy of Science 25 (September
1974), pp. 271-84.
12. Examples of such an approach in international relations might include Graham Allision,
Essence of Decision (Boston: Little Brown, 1971) and John Steinbrunner, The Cybernetic
Theory of Decision (Princeton: Princeton University Press, 1974).
13. This is probably the most persistently cited problem in the individualist program of
reducing all social scientific explanations to the properties of individuals or their interactions.
See, for example, Maurice Mandelbaum, "Societal Facts," British Journal of Sociology 6
(1955); Steven Lukes, "Methodological Individualism Reconsidered," British Journal of
Sociology 19 (June 1968), pp. 119-29; Harold Kincaid, "Reduction, Explanation, and Individ-
ualism," Philosophy of Science 53 (December 1986), pp. 492-513.
talist" except in terms of the individual's social relations to other agents.

The second approach, then, is to explicate and theorize these relationally or
structurally defined properties, that is, to develop a social theory of the
state. Yet these are precisely the properties that neorealism abstracts away
from when it reduces system structures to a distribution of properties of
preexisting individuals. The neorealist's individualist conceptualization of
system structure is therefore too weak to support a social theory of the state:
system structures cannot generate agents if they are defined exclusively in
terms of those agents in the first place. The consequence of making the
individual ontologically primitive, in other words, is that the social relations
in virtue of which that individual is a particular kind of agent with particular
causal properties must remain forever opaque and untheorized.
One response to this might be to argue, as Waltz does,14 that the develop-
ment of an explicit theory of the state is not integral to the development of
systemic theories of international relations. Yet clearly some model of the
state is necessary to build systemic theories of international relations, and
this model can strongly affect the content of those theories. Thus, to argue
that the structure of the industrialized states' interaction with respect to
international trade is an n-person iterated Prisoner's Dilemma, and that free
trade is therefore always problematic, requires a certain understanding of
those states and their interests and powers. The issue, then, is not whether
some understanding of the state is necessary to build systemic theories (it
is), but whether that understanding follows from a theory, grounded in a
coherent set of propositions with some correspondence to reality, or simply
from a set of pre-theoretical assumptions, grounded in intuition or ideology.
Whatever its advantages in terms of analytical convenience, a reliance on
untheorized assumptions about primitive terms leaves us unable to justify
particular conceptualizations of interaction situations and leads, therefore,
to an untenable "as i f approach to systemic theory building.15 Thus, with-
out an explicit theory of the state's powers and interests in international
trade, without a theory of the "rules of the game," it cannot be determined
whether or not this game really is a Prisoner's Dilemma rather than, as some
14. "Reflections on Theory of International Politics: A Response to My Critics," in Robert

Keohane, ed., Neorealism and Its Critics (New York: Columbia University Press, 1986), p.
340.
15. The debate over the validity of theories built on the assumption that the social world
operates "as if" certain things were true is a long one, so my saying that such reasoning is
"untenable" is, of course, contentious. The terms of the debate were first defined by Milton
Friedman's "The Methodology of Positive Economics," in his Essays in Positive Economics
(Chicago: Chicago University Press, 1953), a piece which initiated a lively debate with Paul
Samuelson and others in the pages of the American Economic Review in the early 1960s. For a
particularly cogent argument that "as if" reasoning is inconsistent even with the logical empiri-
cist conception of scientific explanation that informed Friedman's seminal contribution, see
Terry Moe, "On the Scientific Status of Rational Models," American Journal of Political
Science 23 (February 1979), pp. 215-43.
neo-Marxists might argue, a pure cooperation game.16 And, without a com-

pelling argument to that effect, we cannot know if a failed prediction is due
to an error in our systemic theory or to a misspecification of the (pretheoret-
ical) rules of the game. Without a theory of the state, in other words,
neorealists' efforts to build compelling systemic theories of international
relations are seriously compromised.
This consequence does not mean that the actual, micro-economic, as-
sumptions neorealists make about the state are wrong or misleading (al-
though I think they probably are)—just that they lack a theoretical
foundation with some demonstrated correspondence to reality. The result is
a tendency to argue that states act "as i f they maximize, for example,
power and wealth, and a corresponding inability to build credible systemic
theories of international relations. Waltz's separation of theories of the state
and of international relations notwithstanding, if neorealists want to avoid
these problems they must ultimately develop a social theory of the state, that
is, they must make the state theoretically "problematic." This would re-
quire an attempt to theorize directly about the generative structures of the
world and domestic political-economy which constitute states as particular
kinds of agents with certain causal powers and interests. Neorealists already
have an implicit theory of these social relations (if they did not, they could
not attribute any causal powers or interests to state agents), but they cannot
make this theory explicit, and therefore falsifiable, as long as they treat the
state as ontologically primitive. Since the social relations which constitute
states as states will be potentially unobservable and irreducible to the prop-
erties of states themselves, however, such a theoretical reorientation will
require a non-individualist and non-empiricist understanding of system
structures and structural analysis, an understanding of structure as some-
thing more than a distribution of capabilities.
b. World-system theory
World-system theory offers such an understanding of structure and thus, at
least with respect to its conceptualization of structure and structural analy-
sis, can be seen as a progressive problem shift over neorealism. In one
crucial respect, however, world-system theorists duplicate the neorealist
16. As far as I know, no neo-Marxist has used game-theoretic language to characterize
international economic relations between the advanced industrialized countries. But clearly,
because of their very different theoretical understanding of the state, neo-Marxist scholars are
much less likely than neorealists to see those relations in mercantilist, and therefore politically
fragile, terms; see, for example, Robin Murray, "The Internationalization of Capital and the
Nation-State," New Left Review 67 (May-June 1971), pp. 84-109, and John Willoughby, "The
Changing Role of Protection in the World Economy," Cambridge Journal of Economics 6 (June
1982), pp. 195-211. The issue in this article, of course, is not which view is actually correct, but
rather how to develop an approach to the agent-structure problem which ensures at least the
possibility of determining which is correct, that is, of developing a theory of states in interna-
tional economic structures.
approach to the agent-structure problem: they at least implicitly make one

entity primitive, in this case the structure of the world system, and then try
to reduce other entities, such as state and class agents, to its effects. The
result of this strategy, I shall argue, is that world-system theorists reify the
structure of the world system and thus, like neorealists, are unable even in
principle to explain its essential properties. A social structure is reified when
"it is treated as an object analytically independent of the actions by which it
is produced."17 A solution to the agent-structure problem, then, engages in
reification when it objectifies social structures without recognizing that only
human action instantiates, reproduces, and transforms those structures. I
should emphasize that reification presupposes at least an implicit conception
of the relationship of agents to social structures: structures have reproduc-
tive requirements which, for whatever reason, agents passively implement.
The problem with reification, therefore, does not concern the inclusion or
exclusion of agents per se from social scientific theories (since they must be
included), but rather the terms of their inclusion into those theories.
Immanuel Wallerstein's solution to the agent-structure problem has the
same general form, and thus the same strengths and weaknesses, as Louis
Althusser's structural Marxist solution.18 Like Althusser's insistence on the
"absolute ontological priority of the whole over the parts," 19 the core of
Wallerstein's approach is the proposition that the only meaningful unit of
analysis in comparative or international political economy is the whole
world system. Moreover, both scholars accept the concept and discourse of
"totality," 20 of social wholes that are irreducible, even by composition laws,
to their constituent elements. As a result, Wallerstein, like Althusser, con-
ceptualizes "structure" in structuralist or generative terms rather than in
terms of the observable relations between, or properties of, primitive indi-
viduals. That is, in contrast to the neorealist definition of system structure as
17. Douglas Maynard and Thomas Wilson, "On the Reification of Social Structure," in Scott
McNall and Gary Howe, eds., Current Perspectives in Social Theory, vol. 1 (Greenwich,
Conn.: JAI Press, 1980), p. 287.
18. The structural Marxist approach to the agent-structure problem is discussed in Louis
Althusser and Etienne Balibar, Reading Capital (London: New Left Books, 1970), pp. 180-81,
and in Steven Smith, Reading Althusser (Ithaca: Cornell University Press, 1984), pp. 192-200.
It should be noted, however, that despite the similarities between world-system theory and
structural Marxism with respect to their understandings of the agent-structure relationship, they
differ in important ways on other issues, such as the conceptualization of the capitalist mode of
production. See, for example, Gary Howe and Alan Sica, "Political Economy, Imperialism,
and the Problem of World-System Theory," in McNall and Howe, Current Perspectives in
Social Theory, pp. 235-86.
19. Smith, Reading Althusser, p. 177.
20. They disagree, however, about the exact meaning of this term, that is, about whether
totalities are "expressive" or "structured." On these differences, see Michael Burawoy, "Con-
temporary Currents in Marxist Theory," in Scott McNall, ed., Theoretical Perspectives in
Sociology (New York: St. Martins, 1979), pp. 16-39, and Harvey Kaye, "Totality: Its Applica-
tion to Historical and Social Analysis by Wallerstein and Genovese," Historical Reflections 6
(Winter 1979), pp. 405-19.
the distribution of capabilities across preexisting states, world-system theo-

rists define the structure of the world system in terms of the underlying
organizing principles of the world economy, and in particular of the interna-
tional division of labor, which constitute or generate state and class agents.21
The existence and identity of agents as agents, and therefore of their causal
powers and real interests, is produced, and therefore explained, by their
relation to the totality of the capitalist world system. Thus, state agents are
effects of the structure of the world system in much the same sense that
capitalists are effects of the structure of the capitalist mode of production, or
slaves are effects of the structure of master-slave relationships.
This generative reading of world-system theory presupposes an ontolog-
ical and explanatory distinction between "internal" and "external" rela-
tions.22 Internal relations are necessary relationships between entities in the
sense that the entities depend upon the relation for their very identity. Stan-
dard examples of internal relations are parent-child and master-slave;
neither entity is conceivable without the existence of the other. This implies
that an internal relation cannot be reduced to the properties or interactions
of its member elements; on the contrary, the relationship itself explains
essential properties of each entity, and thus the character of their interac-
tion. External relations, on the other hand, are contingent relationships or
interactions between entities, each of which can exist without the other. The
fact that two states go to war or sign a peace, for example, is not essential to
their identity as states. External relations are important for explaining what
happens to entities in the course of their interaction, but they do not explain
the essential characteristics of those entities themselves.
Generative structures are sets of internal relations. To adopt a generative
approach to theorizing about the structure of the international system, there-
fore, is to understand the state as an effect of its internal relations to other
states and social formations in the world political-economy, rather than
purely as an untheorized cause of international events. The strength of a
generative approach to structural theorizing, then, is that in contrast to
21. My generative reading of world-system theory's conceptualization of structure is charac-

teristic only of the "qualitative" (and at this point, apparently the minority) school of world-
system theorists represented, for example, by Wallerstein and Terence Hopkins. Actually, the
recent debate between qualitative and quantitative world-system theorists is an interesting
example of a quite explicit tension within a single research community between scientific realist
and empiricist conceptions of the ontology and methodology of structural analysis. On this
debate see, for example, Richard Little, "The Systems Approach," in Steve Smith, ed., Inter-
national Relations, British and American Perspectives (Oxford, U.K.: Blackwell, 1985), pp.
71-91, and Peter Taylor, "The Poverty of International Comparisons: Some Methodological
Lessons from World-Systems Analysis" (Department of Geography, University of Newcastle-
upon-Tyne, 1985).
22. Useful discussions of this distinction include Bertell Oilman, Alienation: Marx's Concep-
tion of Man in Capitalist Society (Cambridge: Cambridge University Press, 1976), pp. 26-40,
and Bhaskar, The Possibility of Naturalism, pp. 53-55.
neorealism's individualist approach it is able, in principle, to explain the

causal powers and interests of state and class agents, to make these theoreti-
cally and empirically problematic.
A generative approach to structural analysis does not, however, require
that system structures be reified. World-system theorists begin to reify so-
cial structures when they assert, or imply in their concrete research, not
only that certain social relations are irreducible and constitute the state and
class agents which are their elements, but that these relations are analyt-
ically independent of, and ontologically prior to, those agents. Such a view is
implied by the tendency of world-system theorists to follow Althusser in
treating state and class agents as no more than passive "bearers" of sys-
temic imperatives, a tendency which manifests itself in a reliance on func-
tional explanations of state behavior.23 Functionalism is evident, for
example, in world-system theorists' explanation of general wars directly in
terms of the reproduction requirements of the world-system,24 requirements
which become translated (or translate themselves) into bellicose state inter-
ests, as well as in the interpretation of the rise of socialist states in such a
way that it is consistent with the reproduction requirements of the world
system.25 This is not to say that world-system theorists would consciously
argue that the reproduction of the world-system occurs without state
agency—they probably would not. But nonetheless in many explanations
the world system in effect seems to call forth its own reproduction by states;
this suggests that at least in their concrete research, world-system theorists
treat the world-system as at some level operating independently of state
action, that in practice they reify the world-system. While this result may be
unintended, I do not think it is accidental. On the contrary, it follows inevita-
bly from the fundamental premise of Wallerstein's holism—that the whole is
ontologically prior to its parts.
The principal weakness of a structuralist solution to the agent-structure
problem is that, because it cannot "explain anything but behavioral con-
formity to structural demands," 26 it ulthnately fails to provide a basis for
explaining the properties of deep structures themselves. It may be, for ex-
ample, that the division of the world system into three distinct structural
23. This tendency is one of the most persistently cited criticisms of at least the early work in
world-system theory. See, for example, Robert Duplessis, "From Demesne to World-System:
A Critical Review of the Literature on the Transition from Feudalism to Capitalism," Radical
History Review 3 (Fall 1976), pp. 3-41, and Skocpol, "Wallerstein's World Capitalist System."
24. Christopher Chase-Dunn and Joan Sokolovsky, "Interstate Systems, World-Empires
and the Capitalist World-Economy: A Response to Thompson," International Studies Quar-
terly 27 (September 1983), pp. 357-67.
25. Christopher Chase-Dunn, "Socialist States in the Capitalist World-Economy," in his
Socialist States in the World-System (Beverly Hills: Sage Publications, 1982), pp. 21-56.
26. Derek Layder, "Problems in Accounting for the Individual in Marxist-Rationalist Theo-
retical Discourse," British Journal of Sociology 30 (June 1979), p. 150.
positions (core, semi-periphery, periphery) is functional for the reproduction

of capitalism, but this does not explain why the system developed that
particular structure, nor does it guarantee that that structure will endure.27
Because of their passive conception of state and class agency, world-system
theorists tend to fall into an historical determinism which, by ignoring other
possible historical trajectories, implicitly assumes that the evolution of the
world system could not have turned out any other way. The limitations of
world-system theory's solution to the agent-structure problem, and also its
similarities to that of structural Marxism, are in this respect no more appar-
ent than in Wallerstein's explanation of a fundamental structural change like
the transition from feudalism to capitalism.28 Without a recognition of the
ontological dependence of system structures on state and class agents, Wal-
lerstein is forced into an explanation of that transition in terms of exogenous
shocks and the teleological imperatives of an immanent capitalist mode of
production. This explanation opposes Robert Brenner's more or less struc-
turationist one, which describes the transition in terms of a dialectic of
endogenous class struggle and structural conditioning.29 This explanation
reflects difficulties remarkably similar to those structural Marxists have had
in theorizing the transition from one mode of production to another.30
World-system theorists, then, like neorealists, treat their primitive units,
in this case the structure of the world system, as given and unproblematic.
This treatment leads them to separate the operation of system structures
from the activities of state and class agents—in other words, to reify system
structures in a way which leads to static and even functional explanations of
state action. The world system is not treated as an historically contingent,
and therefore continuously problematic, creation and recreation of state and
class agents. I think the greater attention Wallerstein's later work gives to
problems of agency indicates an awareness of this difficulty,31 and these
efforts have helped to move world-system theory away from the excessive
functionalism evident in his early contributions and, perhaps, in my por-
27. Emile Durkheim makes exactly this point in The Rules of Sociological Method (Chicago:
Chicago University Press, 1938), p. 90, when he says that "to show how a fact is useful is not to
explain how it originated or why it is what it is. The uses which it serves presuppose the specific
properties characterizing it, but do not create them. The need we have of things cannot give
them existence, nor can it confer their specific nature upon them. It is to causes of another sort
that they owe their existence."
28. See, for example, Wallerstein, The Modern World-System I (New York: Academic Press,
1974), especially chap. 1.
29. Brenner, "The Origins of Capitalist Development: A Critique of Neo-Smithian Marx-
ism," New Left Review 104 (July-August 1977), pp. 25-92.
30. See, for example, Stephen Resnick and Richard Wolff, "The Theory of Transitional
Conjunctures and the Transition from Feudalism to Capitalism," Review of Radical Political
Economics 11 (Fall 1979), pp. 3-22, and the response in the same issue by Herbert Gintis, "On
the Theory of Transitional Conjunctures," pp. 23-31.
31. See, for example, I. Wallerstein, The Politics of the World-Economy (Cambridge: Cam-
bridge University Press, 1984), pp. 112-46.
trayal of world-system theory. Indeed, in contrast to the complete neglect of

generative structures by neorealists, world-system theorists seem to be
aware of the need to arrive at a solution to the agent-structure problem that
integrates generative structures and state and class agency. World-system
theory's evolution towards a greater focus on agency thus, in some ways,
parallels the development of structural Marxism in the later work of Nicos
Poulantzas.32 As with Poulantzas, however, it is difficult to see how agency
can be brought on an ontological and explanatory par with system structures
without explicitly jettisoning the strict holism and structural reification
which characterize the structuralist approach to the agent-structure prob-
lem. Nonetheless, if only because it at least recognizes the existence and
explanatory role of generative structures, the world-system solution to the
agent-structure problem is considerably closer to that of structuration theory
than is the neorealist solution.
c. Summary
In this section, I have tried to identify important differences between the
neorealist and world-system understandings of "structural" explanation,
and to link these differences to their different social ontologies. I have also
attempted to show that, despite these differences, neorealism and world-
system theory share a common, underlying approach to the agent-structure
problem: they both attempt to make either agents or structures into primitive
units, which leaves each equally unable to explain the properties of those
units, and therefore to justify its theoretical and explanatory claims about
state action. The obvious implication of this argument is that neither state
agents nor the domestic and international system structures which constitute
them should be treated always as given or primitive units; theories of inter-
national relations should be capable of providing explanatory leverage on
both. This does not mean that a particular research endeavor cannot take
some things as primitive: scientific practice has to start somewhere. It does
mean, however, that what is primitive in one research endeavor must be at
least potentially problematic (or function as a "dependent variable") in
another—that scientists need theories of their primitive units. Notwith-
standing their apparent aspiration to be general theories of international
relations, the individualist and structuralist ontologies of neorealism and
world-system theory preclude the development of such theories. In contrast,
a structurationist approach to the agent-structure problem would permit us
to develop theoretical accounts of both state agents and system structures
without engaging in either ontological reductionism or reification.
32. State, Power, Socialism (London: Verso, 1978); see also Bob Jessop, Nicos Poulantzas:
Marxist Theory and Political Strategy (New York: St. Martins, 1985).
3. An alternative approach to the agent-structure problem
Structuration theory is a relational solution to the agent-structure problem

that conceptualizes agents and structures as mutually constituted or co-
determined entities. What this means will become more apparent later, but
first I want to consider the philosophical foundations of structuration theory in
scientific realism, and the current debate of the latter with empiricism in the
philosophy of science. This digression is important to my argument because
of the continuing hegemony of empiricist discourse on social scientists'
conceptualization of "science," and the real possibility that skeptics might
use that discourse to write off structuration theory's generative approach to
structural theorizing as "metaphysical."33 In contrast to empiricism, scien-
tific realism can, in principle, call an ontology "scientific" even if it includes
unobservable generative structures. While scientific realism does not man-
date such an ontology for social life (or, for that matter, any other solution to
the agent-structure problem), it is a necessary condition for the ontology of
structuration theory.
a. Scientific realism
The philosophy of science community is currently in the midst of a wide-
ranging debate between empiricists and scientific realists about what might
be called the "theory of science."34 At issue in the debate are fundamental
questions of ontology, epistemology, and the rational justification of re-
search practices in both the natural and social sciences. Rather than attempt
to review the entire debate, I will concentrate on contrasting the "hard
33. This kind of dismissal is an old individualist move; see, for example, May Brodbeck's
juxtaposition of methodological individualism with "metaphysical" holism in her "Method-
ological Individualisms: Definition and Reduction," in O'Neill, Modes of Individualism and
Collectivism, pp. 289-90. More recently, "analytical Marxists" have resurrected this argument
to motivate a reconstruction of Marxist theory on "micro-foundations"; see Jon Elster, Making
Sense of Marx (Cambridge: Cambridge University Press, 1985), pp. 3-8. In this latter context, it
is perhaps worth noting that a number of social scientific realists have argued that Marxist
theory is best understood in realist, rather than empiricist, terms and therefore does not need to
be reconstructed on microfoundations to be "scientific"; see Russell Keat and John Urry,
Social Theory as Science (London: Routledge & Kegan Paul, 1982), pp. 96-118, and James
Farr, "Marx's Laws," Political Studies 34 (June 1986), pp. 202-22.
34. The terms "empiricist" and "scientific realist" are the labels the participants in this
debate, most of whom are philosophers of natural science, use to describe themselves. Some of
the important contributions and overviews are Hilary Putnam, Mathematics, Matter, and
Method (Cambridge: Cambridge University Press, 1975); Bas van Fraassen, The Scientific
Image (Oxford: Clarendon Press, 1980); Ian Hacking, Representing and Intervening (Cam-
bridge: Cambridge University Press, 1983); Richard Boyd, "On the Current Status of the Issue
of Scientific Realism," Erkenntnis 19 (May 1983), pp. 45-90; Jerrold Aronson, A Realist Philos-
ophy of Science (New York: St. Martins, 1984); Jarrett Leplin, ed., Scientific Realism (Berke-
ley: University of California Press, 1984); Wesley Salmon, Scientific Explanation and the
Causal Structure of the World (Princeton: Princeton University Press, 1984); and Paul Church-
land and Clifford Hooker, eds., Images of Science: Essays on Realism and Empiricism
(Chicago: Chicago University Press, 1985).
Agent-Siructure problem 351
core" empiricist and realist positions on two issues that are relevant to the
agent-structure problem and to structuration theory in particular: 1) the
legitimacy of ascribing ontological status to unobservable entities like
generative structures, and 2) the nature of causal claims and scientific expla-
nation. If the realist positions on these issues seem upon reading to be
unexceptionable, that is because they are: one of the principal arguments for
scientific realism is that it claims to make better sense than empiricism of the
actual research practices of natural and, to a lesser extent, social scientists.
In other words, realists assume that scientists, not philosophers, are the final
arbiters of what is "scientific." This contrasts with the empiricist position,
which is quite explicitly an artificial reconstruction of what scientists are or
should be doing. Indeed, it could be argued that neorealists and world-
system theorists are, at least in some respects, "closet" scientific realists.35
The explicit metatheoretical statements of both remain within an empiricist
discourse,36 however, and thus their research practice does not follow
through on the methodological implications of the scientific realist model.
This failure suggests an ironic twist on the old behavioral argument that the
social sciences are "immature" because they are not "scientific" enough: a
realist might argue that, far from being part of the solution, the empiricist
conception of natural science upon which mainstream social science is based
is part of the cause of its theoretical impoverishment.
The first axis of debate between empiricists and realists is the ontological
status of unobservables. Empiricists tend to "equate the real with the expe-
rientially knowable" in the sense that they are unwilling to say that entities
exist if we cannot, at least in principle, have direct sensory experience of
them. They argue that we should remain, at most, agnostic about the exis-
tence of unobservable entities like quarks, utilities, or generative structures,
and that we should instead interpret the theoretical terms describing such
entities, and the theories in which those terms are embedded, "instrumen-
tally" rather than "realistically."37 Theories and theoretical terms are useful
35. Neorealists might be seen as scientific realists to the extent that they believe that state
interests or utilities are real but unobservable mechanisms which generate state behavior, while
world-system theorists would be realists to the extent that they believe that the structure of the
world-system is a real but unobservable entity which generates agents.
36. The most explicit recent discussion of the philosophy of science underlying neorealism of
which I am aware is the symposium around Bruce Bueno de Mesquita's "Toward a Scientific
Understanding of International Conflict: A Personal View," International Studies Quarterly 29
(June 1985), pp. 121-36. Bueno de Mesquita's emphasis on deductive analysis and logical proof,
rather than the identification of potentially unobservable causal mechanisms, as the foundation
of scientific explanation displays a clearly empiricist epistemological orientation. The explicit
statements on philosophy of science by at least the quantitative school of world-system theo-
rists show a similar reliance on empiricist arguments; see, for example, Christopher Chase-
Dunn, "The Kernel of the Capitalist World-Economy: Three Approaches," in Thompson, ed.,
Contending Approaches, pp. 55-78.
37. The best recent defense of instrumentalism and empiricism more generally is van Fraas-
sen, The Scientific Image.
instruments for organizing and predicting experience, but there is too much
epistemic risk of positing false entities to justify an "abductive inference,"38
an inference that theoretical terms refer to real but unobservable entities and
processes. Empiricists, then, in effect subordinate ontology to epistemol-
ogy—what exists is a function of what can be known experientially.39 In
contrast to empiricists' rejection of abductive inference, scientific realists
argue that such inferences are, in principle, justified if the entity in question
can produce observable effects,40 or if its manipulation permits us to inter-
vene with effect in the observable world.41 Thus, the fact that we can use
theories about the (unobservable) internal structure of atoms to build nu-
clear weapons which can destroy cities is a good reason for the realist to
believe that such structures exist, as we understand them today. This thesis
is important to structuration theory because, in contrast to empiricism, talk
of unobservable and irreducible social structures can be scientifically legiti-
mate in this view. As long as they have observable effects or are manipulable
by human agents, we can, in principle, speak meaningfully about the "real-
ity" of unobservable social structures. "Generative structure," in other
words, is a (potentially) scientific rather than metaphysical concept.
Scientific realists commonly adduce two basic arguments in favor of ab-
ductive inference and, more generally, of the ontological status of unobserv-
ables.42 Proponents of the "indispensability argument" argue that a realist
construal of theoretical terms is necessary to make sense of the actual re-
search practices of natural and social scientists.43 Physicists would not posit
and build tests around quarks, and social scientists would not posit and build
tests around utilities or modes of production, if they thought that these
entities, despite being unobservable, were not real and causally efficacious.
Proponents of the "miracle argument," in turn, go one step further by
arguing that not only is scientific realism necessary to make scientific prac-
tices rationally intelligible, but it is also necessary to explain the instrumen-
38. Abduction is also known as "retroduction." Useful discussions of abduction are found in
Norwood Hanson, "Retroduction and the Logic of Scientific Discovery," in Leonard Krimer-
man, ed., The Nature and Scope of Social Science (New York: Appleton-Century-Crofts,
1969), pp. 73-83, and Boyd, "On the Current Status of Scientific Realism," especially pp. 72-
89. An unusually detailed and explicit illustration of abductive reasoning in the social sciences
(and thus supporting my earlier suggestion that some social scientists are practicing scientific
realists) is found in Elinor Ostrom's "An Agenda for the Study of Institutions," Public Choice
48 (no. 1, 1986), p. 19.
39. Aronson, A Realist Philosophy of Science, p. 261.
40. Bhaskar, The Possibility of Naturalism, p. 16.
41. Hacking, Representing and Intervening; Thomas Cook and Donald Campbell, "The
Causal Assumptions of Quasi-Experimental Practice," Synthese 68 (July 1986), especially pp.
169-72.
42. Alison Wylie, "Arguments for Scientific Realism: The Ascending Spiral," American
Philosophical Quarterly 23 (July 1986), pp. 287-97.
43. Bhaskar, The Possibility of Naturalism, p. 22. Geoffrey Hellman, "Realist Principles,"
Philosophy of Science 50 (June 1983), especially pp. 231-32.
tal success of science in controlling the world.44 If mature scientific theories

did not at least partially correspond to the deep structure of reality, the
success of science would be an unexplainable "miracle." Certainly both of
these arguments depend on "mature theories" for their force, and thus
might be less convincing in the social than natural scientific context (al-
though micro-economic theory and Marxist theory might be candidates for
such a status). But the relative weakness of extant social scientific theories
does not jeopardize the scientific realist's basic point: it is a well-established
and perfectly legitimate scientific practice to posit unobservable entities to
account for observable behavior. Acceptance of this practice does not imply
that any posit is a good one; scientists must still adduce direct or indirect
evidence for the validity of their ontological claims, and this evidence is
always revisable. But by the same token, scientists, not philosophers of
science, are the arbiters of that evidence.
The differences between empiricists and scientific realists over ontology
fuel debate on a second axis, the nature and requirements of scientific
explanation. Traditionally, there have been two competing ideals of scien-
tific explanation: the empiricist or "nomothetic" view that explanation in-
volves the subsumption of a phenomenon under a lawlike regularity; and the
realist or "retroductive" view that it involves the identification of the under-
lying causal mechanisms which physically generated the phenomenon.45 The
resurgence in recent years of the scientific realist view coincides with re-
newed attacks on the Humean model of causation, which supports the em-
piricist account of explanation.46 In the Humean model, a causal relation is a
"constant conjunction" of temporally sequenced observed events that
stands in a relation of logical necessity to certain initial conditions and
laws.47 On the Humean view, our inability to experience causal mechanisms
directly prevents us from imputing any natural necessity to causal relations.
Scientific realists criticize this model because constant conjunctions and
generalizations are not their own explanation, and argue instead that to
44. See, for example, Putnam, Matter, Mathematics, and Method; Boyd, "On the Current
Status of the Issue of Scientific Realism"; Richard Schlagel, "A Reasonable Reply to Hume's
Skepticism," British Journal for the Philosophy of Science 35 (December 1984), pp. 359-74.
45. See Ernan McMullin, "Two Ideals of Explanation in Natural Science," in Peter French,
et al., eds., Causation and Causal Theories (Minneapolis: University of Minnesota Press,
1984), pp. 205-20, and the three-way debate between Philip Kitcher, Bas van Fraassen, and
Wesley Salmon in "Approaches to Explanation," The Journal of Philosophy 82 (November
1985), pp. 632-54.
46. Rom Harre and Edward Madden, Causal Powers (Totowa, N.J.: Rowman and Littlefield,
1975); Salmon, Scientific Explanation; Schlagel, "Hume's Skepticism."
47. Hence behavioral social scientists' emphasis on quantitative analysis to discover law-like
regularities, rather than qualitative analysis and theory to identify causal mechanisms. On the
empiricist model, we cannot have science without (relatively) "constant" conjunctions. For a
useful more or less realist critique of this model of causation as it relates to social science, see
Daniel Hausman, "Are There Causal Relations among Dependent Variables?" Philosophy of
Science 50 (March 1983), pp. 58-81.
make a true explanatory claim, it is necessary to identify the underlying

causal mechanisms which make an event naturally necessary. The disagree-
ment here ultimately hinges on the legitimacy of abductive inference. Real-
ists argue that if we can explain the physical dispositions and causal powers
Of unobservable entities, we can make a legitimate abductive inference
about the existence of naturally necessary relations between cause and ef-
fect, and thereby transcend Hume's skepticism about causation.48 In effect,
the realist's focus on causal mechanisms is an attempt to explain the empiri-
cist's constant conjunctions. These different models of causation, then, gen-
erate very different models of scientific explanation. Whereas the empiricist
explains by generalizing about observable behavior, the realist explains by
showing how the (often unobservable) causal mechanisms which make ob-
servable regularities possible work. The scientific realist, in other words,
argues that "answers to why-questions require answers to how- and what-
questions." 49
I want to conclude this brief overview of the empiricist/realist debate with
three points about its implications for structuration theory and, more gener-
ally, for social scientific research practice. First, scientific realism attempts
to make sense of what practicing natural and social scientists in fact do,
rather than prescribing on the legitimacy of certain research practices versus
others. Abduction and the de facto positing of unobservable causal pro-
cesses and entities, whether those are quarks, utilities, or modes of produc-
tion, go on constantly in scientific research, and scientific realists see no
reason to write this practice off as unscientific. For this reason, and this is
the second point, in contrast to empiricism, scientific realism can make
scientific sense of unobservable generative structures, of structures that are
irreducible to and generate their elements. This outflanks a key motivation
for individualism, namely that structural theorizing of the generative variety
is necessarily "metaphysical" or "unscientific." Finally, although there are
important problems in translating the protocols and discourse of natural
scientific practice directly to the social sciences—what Roy Bhaskar calls
"limits to naturalism"50—the basic realist idea that scientific explanation
48. Some realist accounts of causation, and particularly the account of Harre and Madden,
have been accused of implying an Aristotelian "essentialism"—the explanation of observable
phenomena in terms of occult and impenetrable "essences"; see, for example, David Miller,
"Back to Aristotle," British Journal for the Philosophy of Science 23 (February 1972), pp. 69-
78, and Fred Wilson, "Harre and Madden on Analyzing Dispositional Concepts," Philosophy
of Science 52 (December 1985), pp. 591-607. Other realists, however, emphasize that this
objection can be vitiated by explaining causal powers in terms of the physical properties and
social relations which underlie them; Schlagel, "Hume's Skepticism."
49. Russell Keat and John Urry, Social Theory as Science (London: Routledge & Kegan
Paul, 1982), p. 31.
50. Perhaps the most difficult problems in making this translation concern the role of human
motivations and self-understandings in social scientific explanations, and the ambiguity of the
notion of causal "mechanisms" in social life. For a sample of the recent debate among scientific
realists on the limits of naturalism in the social sciences, see Bhaskar, The Possibility of
Naturalism, and Keat and Urry, Social Theory as Science, especially the postscript.
consists in the identification of underlying causal mechanisms rather than in

generalizations about observable regularities does apply to the social sci-
ences, and its adoption there would have important implications for the
explanation of social action. Moreover, and perhaps more provocatively,
this idea suggests that social scientific practice must be "critical" to be
"scientific." The rest of this paper begins to consider the nature and implica-
tions of this point within a specifically structurationist perspective on the
agent-structure problem.
b. Structuration theory
Scientific realism provides a philosophical basis for a generative approach
to structural theorizing in the social sciences, and in so doing, it provides a
foundation for working out the implications of one of the intuitions about
social life with which I opened the discussion of the agent-structure prob-
lem: that the capacities and even existence of human agents are in some way
necessarily related to a social structural context—that they are inseparable
from human sociality. The implications of this insight, however, can be
worked out in at least two different ways—one reifies the social relations
that constitute agents, and one does not. I argued earlier that world-system
theory embodies a structuralist approach to the agent-structure problem that
is prone to reification and determinism. Structuration theory attempts to
preserve the generative and relational aspects of structuralism while taking
explicit conceptual and methodological steps to prevent the analytical sep-
aration of generative structures from the self-understandings and practices
of human agents to prevent structural reification.
It may be useful to preface the discussion with some comments on what
structuration theory, as a theory, is about. Structuration theory is an "ana-
lytical" rather than "substantive" theory, in the sense that it is about the
analysis rather than the substance of the social world.51 Structuration theory
says something about what kinds of entities there are in the social world and
how their relationship should be conceptualized, and as such it provides a
conceptual framework or meta-theory for thinking about real world social
systems, but it does not tell us what particular kinds of agents or what
particular kinds of structures to expect in any given concrete social system.
Structuration theory, then, does not compete directly with neorealism or
world-system theory, but instead with their individualist and structuralist
approaches to the agent-structure problem—that is, with their social on-
tologies. As a social ontology, however, structuration theory does have
51. Ira Cohen makes this particular distinction in "The Status of Structuration Theory: A
Reply to McLennan," Theory, Culture, and Society 3 (no. 1, 1986), pp. 123-34. Nigel Thrift
makes a similar point, arguing that structuration theory is more meta-theory than theory in
"Bear and Mouse or Bear and Tree? Anthony Giddens' Reconstitution of Social Theory,"
Sociology 19 (November 1985), pp. 609-23.
implications for the potential content of substantive theories about real-

world social systems, and for the methodology that social scientists should
use to study those systems. These implications in turn define a research
agenda for social inquiry. I shall suggest in Section 4 how neorealism and
world-system theory relate to this research agenda, and indicate some of
the gaps in contemporary international relations research that emerge from
this comparison.
In keeping with structuration theory's character as a conceptual approach
rather than a substantively-defined theory, the group of "structuration theo-
rists" is also quite diverse and, indeed, may only be recognizable as such
from outside the structurationist problematic. Thrift, for example, identifies
five major social theorists (Philip Abrams, Roy Bhaskar, Pierre Bourdieu,
Anthony Giddens, and Derek Layder) as structurationists, even though most
of these would probably resist the structurationist label (perhaps because it
is Giddens's).52 Despite their internal differences, however, they all share
four basic analytical objectives that can be seen as defining the "hard core"
of the structuration research program.53
1) In opposition to individualists, they accept the reality and explana-
tory importance of irreducible and potentially unobservable social
structures that generate agents.
2) In opposition to structuralists, they oppose functionalism and stress
"the need for a theory of practical reason and consciousness that
can account for human intentionality and motivation."54
3) These oppositions are reconciled by joining agents and structures in
a "dialectical synthesis" that overcomes the subordination of one
to the other, which is characteristic of both individualism and struc-
turalism.55
4) Finally, they argue that social structures are inseparable from spatial
and temporal structures, and that time and space must therefore be
incorporated directly and explicitly into theoretical and concrete so-
cial research.56
52. Thrift, "On the Determination of Social Action in Space and Time," p. 30.
53. Adapted from ibid., pp. 28-32.
54. Ibid., p. 30.
55. This synthesis requires the development of mediating concepts that can link structure
and agency in concrete situations, and as such is probably the key source of disagreement
among structuration theorists. But whether this linkage is established through a "position-
practice system" (Bhaskar), a "habitus" (Bourdieu), or a "system-institution" nexus (Gid-
dens), they all serve the same theoretical function in concrete research, namely, binding agents
and structures into mutually implicating ontological and explanatory roles.
56. This point is more than a ritual admonition for social scientists to be sensitive to the
historical and geographical context of their subjects: substantive "social theories must be about
the time-space constitution of social structures right from the start." (Thrift, "On the Determi-
nation of Social Action," p. 31, italics in original.)
The following discussion elaborates these points by first discussing the na-
ture of social structures, then of agents, and finally of their interrelationship.
My account relies primarily on Bhaskar's work, which of the five theorists
displays the most explicitly scientific realist orientation.57
Structuration theorists start out much like structuralists by defining
"structure" in generative terms as a set of internally related elements.58 The
elements of a social structure could be agents, practices, technologies, ter-
ritories—whatever can be seen as occupying a position within a social or-
ganization. The fact that these elements are internally related means that
they cannot be defined or even conceived independently of their position in
the structure. Thus, in contrast to the neorealist definition of international
system structures as consisting of externally related, preexisting, state
agents, a structurationist approach to the state system would see states in
relational terms as generated or constituted by internal relations of individu-
ation (sovereignty) and, perhaps, penetration (spheres of influence). In other
words, states are not even conceivable as states apart from their position in a
global structure of individuated and penetrated political authorities. The
nature and configuration of the internal relations that comprise a social
structure, in turn, define a set of possible transformations or combinations of
its elements. As a set of possible transformations, social structures are, by
definition, not reducible to the relationships between a structure's elements
that are observed in a given concrete context. Structures make a given
combination or instantiation of elements possible, but they are not ex-
hausted by whatever particular manifestation is actual.
Structuration theorists argue the scientific realist thesis that because so-
cial structures generate agents and their behavior (in the sense that they
make the latter possible), that because social structures have observable
effects, we can potentially claim that they are real entities despite being
possibly unobservable. This thesis raises the issue of when we can legiti-
mately claim that a social structure exists. The key weakness of abductive
inference is the danger of circular reasoning and self-confirmation; we assert
that a structure exists because it has the observed effects which we posited
for the structure in the first place. This weakness is, I think, at the heart of
57. In his Profiles and Critiques in Social Theory (Berkeley: University of California Press,
1982), Giddens indicates (p. 14) that he also accepts a realist conception of science, but his
realism is generally less explicit and thus more attenuated than Bhaskar's. A more important
reason for relying on Bhaskar rather than Giddens, however, is the latter's weaker conception
of social structure as rules and resources rather than as a set of real but unobservable internal
relations, a conception which is arguably ultimately voluntarist in its implications; see for
example, Alex Callinicos, "Anthony Giddens: A Contemporary Critique," Theory and Society
14 (March 1985), pp. 133-66.
58. See, for example, Bhaskar, The Possibility of Naturalism, especially pp. 47-56; Peter
Manicas, "The Concept of Social Structure," Journal for the Theory of Social Behavior 10
(July 1980), pp. 65-82; Keat and Urry, Social Theory as Science, p. 121; Andrew Sayer,
Method in Social Science: A Realist Approach (London: Hutchinson, 1984), pp. 80-87.
the frequently heard complaint of mainstream social scientists, who claim

that Marxist and other structural theories in social science are, in principle,
"non-falsifiable." The appropriate response to this problem is to find evi-
dence for a structure's existence or workings that are independent of the
particular observations from which the structure was abduced,59 and to rec-
ognize and attempt to control for the radical openness of social systems.60
This is, of course, what natural scientific practice is all about: scientists
work backward from an observed phenomenon to a postulated entity or
causal mechanism, and then they try to develop tests in closed systems to
determine independently whether the inference is justified. Clearly such
independent evidence is harder to obtain in the social sciences, but this does
not invalidate the basic idea of using unobservable structures and causal
mechanisms to account for social behavior. Indeed, in view of the individ-
ualist's rejection of generative structures as "metaphysical," it is ironic that
the generative approach is much closer to the conceptualization of structure
in mathematics and the natural sciences than is the individualist's definition
as a distribution of unit-level properties.61
Up to this point, the structurationist understanding of structure is identical
to the structuralist's. Each conceives of structure in combinatorial terms as
an irreducible entity that "generates" its elements and their possible trans-
formations. Structuration theorists diverge from structuralists, however, in
arguing that social structures differ in at least two fundamental respects from
natural structures, and that a recognition of these differences is essential to
avoid the reification of social structures characteristic of structuralism. The
first difference is that "social structures, unlike natural structures, do not
exist independently of the activities they govern." 62 While it may make
sense to say that a natural structure has an existence apart from the behavior
59. Keat and Urry, Social Theory as Science, postscript.
60. An open system is one in which invariant constant conjunctions do not obtain. Although
the complexity and open-endedness of open systems limit the possibilities for decisive tests of
social scientific claims (see Bhaskar, The Possibility of Naturalism, pp. 164-65), this problem
afflicts not only those theories which refer to unobservable entities. For an interesting and
explicitly realist argument about how open systems might, in some cases, be studied in a way
that would permit relatively controlled tests, see Cook and Campbell, "Quasi-Experimental
Practice."
61. On the definition of structure in mathematics see, for example, Marc Barbut, "On the
Meaning of the Word 'Structure' in Mathematics," in M. Lane, ed., Structuralism: A Reader
(London: Jonathan Cape, 1970), Michael Resnick, "Mathematics as a Science of Patterns:
Ontology and Reference," Nous 15 (November 1981), pp. 529-50, and Stewart Shapiro,
"Mathematics and Reality," Philosophy of Science 50 (December 1983), pp. 523-48. Modern
physics, in turn, is based on group theory (the mathematical theory of binary systems), which is
explicitly combinatorial and possibilistic in its view of structure. I should probably note, how-
ever, that although I emphasize this similarity in social and natural scientific conceptions of
structure, I am not saying that social science should be social physics. I am only trying to justify
a certain kind of thinking and explanation in social science by pointing out that it pervades the
natural sciences as well.
62. Bhaskar, The Possibility of Naturalism, pp. 48-49; on the differences between natural
and social structures, see also Giddens, Studies in Social and Political Theory (London: Hutch-
inson, 1977), pp. 118-19.
of its elements, social structures are only instantiated by the practices of

agents. The deep structure of the state system, for example, exists only in
virtue of the recognition of certain rules and the performance of certain
practices by states; if states ceased such recognition of performances, the
state system as presently constituted would automatically disappear. Social
structures, then, are ontologically dependent upon (although they are not
reducible to) their elements in a way that natural structures are not. The
second difference is that "social structures, unlike natural structures, do not
exist independently of the agents' conceptions of what they are doing in their
activity." 63 In other words, social structures have an inherently discursive
dimension in the sense that they are inseparable from the reasons and self-
understandings that agents bring to their actions. This discursive quality
does not mean that social structures are reducible to what agents think they
are doing, since agents may not understand the structural antecedents or
implications of their actions. But it does mean that the existence and opera-
tion of social structures are dependent upon human self-understandings; it
also means that social structures acquire their causal efficacy only through
the medium of practical consciousness and action.
Just as social structures are ontologically dependent upon and therefore
constituted by the practices and self-understandings of agents, the causal
powers and interests of those agents, in their own turn, are constituted and
therefore explained by structures. The structures that constitute agents are
of two distinct kinds: external, or social, structures; and internal, or organi-
zational, structures. Each type explains a distinct set of the causal powers
and interests of agents—social and intrinsic ones, respectively. Thus, all
agents possess three intrinsic capacities or powers in virtue of their internal
organizational structure or "anatomy": 64 1) to have a theoretical under-
standing (however inaccurate) of its activities, in the sense that it could
supply reasons for its behavior; 2) to reflexively monitor and potentially
adapt its behavior; and 3) to make decisions. These causal powers differ-
entiate agents from the non-sapient elements that comprise natural struc-
tures, and to the extent that states can be considered goal-directed units of
action, they can be considered agents by this definition. Internal organiza-
tional structures are also important, however, for explaining the subjectively
perceived interests of agents. Individual and organizational decision-making
pathologies in the state, for example, may be crucial for determining how
social structural or objective imperatives for competent state practice—
a state's "real interests"—translate into subjective interests and actual
performance.65
63. Bhaskar, The Possibility of Naturalism, pp. 48-49.

64. Adapted from Giddens, The Constitution of Society, pp. 5-6.
65. Structuration theorists have yet to tackle in a sustained way the nature and role of
interests in social scientific explanations. Although some of the more materialistically inclined
structuration theorists might reject the explanatory use of interests altogether, I am inclined to
The importance of their internal organizational structures notwithstand-

ing, however, other causal powers and the real interests of agents are depen-
dent upon and thus explainable only by the external or social structural
context in which they are embedded. For example, the causal powers of
capitalists (for example, to invest and disinvest capital, to extract a surplus
from labor) and their real interests (to maximize profits, and so forth) are a
function of the organizing principles of the capitalist mode of production
which define their position and interests as a capitalist's. Similarly, the
causal powers of the state—to maintain control over the resources and
violence potential in a given territory, to act in an international environment
free of legal compulsion, and so on—are conferred upon it by the domestic
and international social structures in virtue of which it is a state in the first
place. These structural relations may be as general as the organizing princi-
ples of the interstate system (for example, sovereignty, penetration) which
constitute states as such, or they may consist of the more localized organiz-
ing principles of concrete international systems, like the balance of power,
which define particular kinds of states. Thus, the "balancer" in a balance-of-
power system, or a core state in the capitalist world economy, has certain
powers, responsibilities, and interests which it possesses only in virtue of its
social structural position.66 Social structures, then, constitute the conditions
of existence of states and state action; indeed, without social structuring
principles one could not talk meaningfully about the fundamental building
blocks of international relations: "states," "state powers," "foreign pol-
icy," and so forth. Put in another way, international and domestic structures
generate the "rules of the game" (broadly defined to include state agents
themselves) within which states interact.
Structuration theory, then, conceptualizes agents and structures as mutu-
ally constitutive yet ontologically distinct entities. Each is in some sense an
effect of the other; they are "co-determined." Social structures are the
result of the intended and unintended consequences of human action, just as
those actions presuppose or are mediated by an irreducible structural con-
text. This understanding of the agent-structure relationship is made possible
by conceptualizing each from the start as ontologically dependent upon the
other, by conceptualizing agents in terms of the internal relations that define
them as such, and by conceptualizing social structures as existing only
through the medium of the agents and practices that they constitute. This is
think that their agent-structure framework presupposes at least an implicit distinction between
"subjective" and "real" interests. The best overview of the various conceptualizations of
"interest," and of the difficulties in explaining interests, is probably still William Connolly's
"Interests in Politics," in his book, The Terms of Political Discourse (Princeton: Princeton
University Press, 1974), pp. 45-84.
66. For a discussion of the balance of power that is consistent in its substance, if not in its
philosophical rationale, with the interpretation I suggest, see Ashley, "The Poverty of Neoreal-
ism," pp. 276-79.
what Giddens means by the "duality of structure," that "the structural

properties of social systems are both the medium and outcome of the prac-
tices which constitute those systems." 67 Structuration theory is therefore
more than an attempt to introduce a greater balance of structure and agency
in social theory than is found in individualism and structuralism. Its social
ontology radically reconceptualizes the fundamental properties of agents
and social structures in such a way to make them ontologically interdepen-
dent, and it is only virtue of this reconceptualization that the "errors" of
reduction and reification characteristic of individualism and structuralism
are avoided. As I shall indicate in the next section, the ontological and
conceptual interdependence of agents and structures in structuration theory
has important implications for the explanation of social action. Put very
generally, it forces us to see agents and structures as simultaneously in-
volved in the production of social phenomena. In Bhaskar's words:
Thus society is not the unconditioned creation of human agency (volun-
tarism), but neither does it exist independently of it (reification). And
individual action neither completely determines (individualism) nor is
completely determined by (determinism) social forms.68
4. Implications for international relations theory
The discussion of structuration theory so far has focused on its social ontol-
ogy, on its conceptualization of the nature and relationship of human or
organizational agents and social structures. While structuration theory does
not by itself generate claims or hypotheses about particular international
system structures or the causes of state action, the realist/structurationist
problematic does have both epistemological and theoretical implications for
the study of international relations. Thus, on the one hand, structuration the-
ory's social ontology strongly conditions its approach to the explanation of
state action. This idea is consistent with the effort of scientific realists to
reverse the subordination of ontology to epistemology, which is characteris-
tic of empiricism, and instead to make the form of scientific explanations
dependent on the nature and causal properties of entities. Beyond this gen-
eral concern with the form of explanations, however, structuration theory
also has implications for the content of substantive international relations
theories or, perhaps more precisely, for the nature and scope of the research
agendas which underlie those theories. In particular, structuration theory
suggests that, while neorealism and world-system theory provide important
insights into the structure and dynamics of international systems, they leave
67. Giddens, Central Problems in Social Theory, p. 69.

68. Roy Bhaskar, "Emergence, Explanation, and Emancipation," in Paul Secord, ed., Ex-
plaining Human Behavior (Beverly Hills: Sage Publications, 1982), p. 286.
important gaps in the theorization of the two basic building blocks of interna-
tional relations theory, states and international system structures.
a. Epistemological implications
Relatively little empirical research has been explicitly informed by struc-
turation theory, which might illustrate its implications for the explanation of
state action.69 In this article, I cannot develop an extended empirical illustra-
ion of my own; instead, I shall adapt materials from the few structuration
theorists70 and realist philosophers of social science who have tackled issues
of social scientific explanation in order to make two general epistemological
arguments: 1) that structural and agent-based analyses have distinct and
irreducible functions in the explanation of social action, but that 2) they are
both necessary elements of a complete explanation of social action. These
two arguments have important implications for our understanding of the
nature and limits of structural and what I shall call historical explanations, as
well as for their integration in "structural-historical" analysis.
Explanations are answers to certain kinds of questions. What counts as an
adequate explanation therefore depends on the object of the question, on
what is taken to be problematic.71 From a structurationist perspective, two
kinds of questions are particularly relevant to the explanation of social ac-
tion: "How is action X possible?" and "Why did X happen rather than Y?"
The domains of these two questions, and therefore the kinds of answers we
would expect, are different. "How-questions" are concerned with the do-
main of the possible, whereas "why-questions" are concerned with the
domain of the actual. To remain clear on the nature and limits of structural
explanation, an explicit epistemological and methodological distinction must
be maintained between the logic of these questions: "structural" analysis
explains the possible, while "historical" analysis explains the actual. His-
torical analysis focuses on what actually happened or will happen, and thus
takes as unproblematic the possibility that those events can happen. Actual
behavior, rather than the range of possible behaviors, is the explanandum.
While this historical analysis of conjunctural causes is an essential component
69. The most extensive use of an explicitly structurationist perspective in empirical research
is probably Allan Pred, Place, Practice, and Structure (Cambridge, U.K.: Polity Press, 1986).
In his Explanation in Social History, however, Lloyd argues (p. 306) that the work of a number
of prestructuration theorists has a distinctly structurationist "structure," including, for ex-
ample, the works of Barrington Moore, Social Origins of Dictatorship and Democracy (Boston:
Beacon Press, 1966), and Alain Touraine, The Self-Production of Society (Chicago: Chicago
University Press, 1977), and Abrams, Historical Sociology.
70. Sayer, Method in Social Science; Sylvan and Glassner, A Rationalist Methodology.
71. The implications of the epistemological distinctions between different kinds of questions
are brought out systematically in Alan Garfinkel, Forms of Explanation (New Haven: Yale
University Press, 1981), especially pp. 21-48. Despite its explicitly anti-realist ontological
perspective, van Fraassen's The Scientific Image is also quite good on the logic or "pragmat-
ics" of different types of explanations.
of the explanation of state action, however, it is only a necessary, not a

sufficient, condition for scientific explanation. Within a scientific realist epis-
temology, why-questions require answers to how-questions; that is, to ex-
plain why a state did X rather Y, we need to know how that state and its
choices were possible in the first place. It is therefore necessary to make the
actual behavior and properties of states and state systems "problematic"
rather than simply accepting them as given. It is necessary, in other words,
to engage in structural analysis to explain the causal properties of states in
virtue of which their actions are possible. The following discussion of struc-
tural and historical analysis expands on the distinct, but nonetheless interde-
pendent, epistemological roles of these two forms of explanation.
Structural research starts with actual events, with history, and by a pro-
cess of critique and abduction—that is, by asking what must exist for those
events to happen—abstracts to the social and internal organizational struc-
tures which make those events possible. These structures might then be
modeled formally with qualitative techniques (mathematical or grammatical)
which describe their possibilistic relationship to observable events,72 al-
though such modeling may not always be possible in open systems. In any
case, structural explanations contribute to the explanation of observable
events by showing that they are instances of the possible ways of acting of
social agents, where those possibilities are defined by the structurally deter-
mined causal powers and interests of those agents. Put another way, struc-
tural explanations reveal the conditions of existence or "rules of the game"
of social action. In this sense structural theory is necessarily "critical,"
since it forces us to look beyond given appearances to the underlying social
relationships that generate (in a possibilistic sense) phenomenal forms.
While structural analyses are thus part of a complete explanation of actual
events, however, they do not explain those events directly; they only an-
swer the question of how they are possible, of what combinations or trans-
formations of a structure's elements are consistent with its organizing
principles. Although structural analyses may uncover "tendencies" for
structures to be actualized in certain ways, neither generalization nor point
prediction is an important aspect of structural explanations, and any attempt
to use them to account directly for the production of particular events would
risk overextending them beyond their proper explanatory domain.73
Historical research, on the other hand, "studies actual events and objects
as 'unities of diverse determinations,' each of which have been isolated and
examined through abstract [structural] research." 74 Historical explana-
72. An excellent introduction to some of the formal methods that could be used in generative
structural analyses is found in chaps. 5 and 6 of Sylvan and Glassner, A Rationalist
Methodology.
73. In Method in Social Science, Sayer argues (p. 217) that a failure to recognize these
limitations of structural analysis is responsible for the deterministic, or what he calls "pseudo-
concrete," quality of much Marxist research.
74. Sayer, Method in Social Science, p. 216.
tions75 take the interests and causal powers of agents as given (or reconstruct
them without trying to explain them), and then attempt to explain particular
events by focusing on how those powers and interests are affected by the
incentives facing actors. Neorealism proceeds at this level; it stipulates the
structural context and the interests and causal powers of agents and then
attempts to answer the question "Why did state X do Y rather than Z?"
Beyond this, however, it is important to note that one of the intended or
unintended effects of state action is to produce or reproduce system struc-
tures; consequently, historical analysis is necessary to explain the emer-
gence and persistence of the structural conditions which constitute the
medium and conditions of possibility for state action. This recursive quality
of structural and historical explanations is the "unity" beneath their "diver-
sity," and thus it is the ultimate basis of their epistemological interdepen-
dence. It is nonetheless necessary to maintain the distinction between and
autonomy of each explanatory mode: each ultimately explains the properties
of the central objects of the other.
The respective explanatory limitations of structural and historical anal-
yses suggest, however, that a complete explanation of state action—that is,
one that explains both how that action was possible and why that possibility
was actualized in a particular form at a given moment—will have to combine
these methodologies into a "structural-historical" or "dialectical" analy-
sis.76 This combination will require abstract structural analysis to theorize
and explain the causal powers, practices, and interests of states, and con-
crete historical analysis to trace the causally significant sequence of choices
and interactions which lead to particular events (and to the reproduction of
social structures). Given the difficulty of doing structural and historical re-
search simultaneously, structural-historical analysis may require "brack-
eting" first one and then the other explanatory mode,77 that is, taking social
75. By my use of the term "historical" to describe this form of explanation, 1 do not mean to
suggest that this is the explanatory mode historians always use, or that the research practice of
historians is necessarily astructural or atheoretical. On the contrary, it seems to me that just as
good social science is historical, good history is structural and theoretical. I am only trying to
argue that "historical" and "structural" explanations are epistemologically distinct but interde-
pendent forms of inquiry, regardless of who uses them.
76. The term "structural-historical" is from Fernando Cardoso and Enzo Faletto, Depen-
dency and Development in Latin America (Berkeley: University of California Press, 1978), pp.
ix-xiv, while "dialectical" is from Sylvan and Glassner, A Rationalist Methodology, pp. 154-
59; both terms parallel the relationship between "abstract" and "concrete" research in Sayer's
Method in Social Science. Although he does not use either of these terms, Peter Manicas
provides a good illustration of the logic and implications of this form of inquiry in his critique of
Theda Skocpol's State and Social Revolutions (Cambridge: Cambridge University Press, 1979);
see his review in History and Theory 20 (no. 2, 1981), pp. 204-18.
77. Giddens, Central Problems in Social Theory, pp. 80-81. This notion of "bracketing" is a
focal point of some of the major critiques of structuration theory; see, for example, Margaret
Archer, "Morphogenesis versus Structuration: On Combining Structure and Action," British
Journal of Sociology 33 (December 1982), pp. 455-83, and Nicky Gregson, "On Duality and
Dualism: The Case of Structuration Theory and Time Geography," Progress in Human Geog-
raphy 10 (June 1986), pp. 184-205.
structures and agents in turn as temporarily given in order to examine the

explanatory effects of the other. This methodological difficulty, however,
should not obscure the epistemological interdependence of structural and
historical analysis, the fact that the respective explanatory roles of agents
and social structures cannot be understood apart from their interrelation-
ship. This conclusion follows directly from the ontology of structuration
theory. Agents are inseparable from social structures in the sense that their
action is possible only in virtue of those structures, and social structures
cannot have causal significance except insofar as they are instantiated by
agents. Social action, then, is "co-determined" by the properties of both
agents and social structures.
b. Theoretical implications
While this discussion of some epistemological implications of structura-
tion theory is admittedly very general, it is nonetheless relevant to the scope
and content of substantive international relations theories. A key implication
of the argument in Section 2 about the agent-structure relationship was that
theories of international relations must have foundations in theories of both
their principal units of analysis (state agents and system structures). Such
theories are more than simply convenient or desirable: they are necessary
to explain state action. This requirement follows directly both from the
scientific realist's conception of explanation as identifying causal mecha-
nisms, and from the ontological claims of structuration theory about the
relationship of agents and structures. If the properties of states and system
structures are both thought to be causally relevant to events in the interna-
tional system, and if those properties are somehow interrelated, then theo-
retical understandings of both those units are necessary to explain state
action. Waltz's suggestion that the theory of the state is not integral to the
task of developing systemic theories of international relations must therefore
be rejected. Structuration theory provides a conceptual and methodological
framework to overcome this separation, and as such it defines a research
agenda for theorizing about both state agents and the system structures in
which they are embedded. The core of this agenda is the use of structural
analysis to theorize the conditions of existence of state agents, and the use of
historical analysis to explain the genesis and reproduction of social struc-
tures. Although even preliminary remarks about the possible content of such
theories would require another article, I can indicate some of the directions
and bodies of research which might be relevant to such a research agenda.
"Theorizing the state" implies a research endeavor which seeks to de-
velop a theoretically and empirically grounded understanding of the causally
significant properties (such as powers, interests, practices) of the state as an
organizational agent or entity. Ideally such a theory would define exhaus-
tively the possible ways of acting of state agents, rather than generate deter-
minate predictions about particular state behaviors. The possible ways of

acting of an agent are constituted and therefore explained by the social
structural context in which it is embedded and by its internal organizational
structure, and as such they are amenable to structural explanation. Thus,
structural analysis can be used to explicate the social structural organizing
principles which generate the state as a particular kind of social actor, that
is, in virtue of which the state is a state in the first place. This use would
recognize the state as an inherently social entity, rather than as a Hobbesian
primitive individual. Structural analysis could also reveal the internal organi-
zational structures of the state which condition its perceptions and re-
sponses to social structural imperatives and opportunities. International
relations theorists have tended to discount the importance of such internal
organizational structures in explaining state action,78 but given that these
structures will be the proximate cause of any state action, they are likely to
constitute important mechanisms in the production of state behavior.
To elaborate a little further, at least four social structures might constitute
states: domestic-economic, domestic-political, international-economic,
and international-political structures.79 Recent work consistent with the
critical-structural orientation of structuration theory has examined all these
structures, although much remains quite preliminary and unintegrated with
other bodies of structural research. The most sustained analyses of the
state's structural conditions of existence are found within the neo-Marxist
tradition and within the debates of the latter with Weberians.80 While the
neo-Marxist literature is concerned primarily with the relationship of the
(capitalist) state to domestic economic structures, world-system theorists
have concentrated on the state's role in international economic structures.
Although the world-system understanding of the structure of global capi-
78. Stephen Krasner, "Are Bureaucracies Important?" Foreign Policy 7 (Summer 1972), pp.
159-79; Robert Art, "Bureaucratic Politics and American Foreign Policy: A Critique," Policy
Sciences 4 (December 1973), pp. 467-90.
79. This multiplicity of structures implies a rejection of what might be called structural
monism, that is, the view that there is only one set of underlying organizing principles, such as
those of the economy, that can be explicated in generative terms and therefore constitutive of
agents. This anti-monism is consistent with the critique of structural Marxism developed by
post-Althusserians like Barry Hindess and Paul Hirst in Mode of Production and Social Forma-
tion (London: MacMillan, 1977), and Ernesto Laclau and Chantal Mouffe in Hegemony and
Socialist Strategy (London: Verso, 1982). But their discourse-theoretic solution to the problem
of structural monism in many ways fundamentally opposes my suggestion that we build theories
of multiple social structures on the basis of scientific realism.
80. Prominent examples of neo-Marxist state theory include John Holloway and Sol Pic-
ciotto, eds., State and Capital: A Marxist Debate (London: Edward Arnold, 1978); Poulantzas,
State, Power, Socialism; and Goran Therborn, What Does the Ruling Class Do When It Rules?
(London: New Left Books, 1978). Weberian critiques include Theda Skocpol, "Political Re-
sponse to Capitalist Crisis: Neo-Marxist Theories of the State and the Case of the New Deal,"
Politics and Society 10 (no. 2, 1981), pp. 155-201, and Michael Mann, "The Autonomous
Power of the State," European Journal of Sociology 25 (no. 2, 1984), pp. 185-213.
talism is arguably marred by an overemphasis on exchange relationships,81

both the neo-Marxist and world-system literature offer important insights
into the economic conditions of existence of the state, and therefore some
of its causal powers and liabilities. Less research, I think, has been done
from a critical-structural perspective on the political structures that might
constitute the state. Nonetheless, at the domestic level, innovative work
is being done which takes as its starting point a Gramscian rejection of the
economism of most neo-Marxist theory and instead attempts to theorize
political forms in critical-structural terms.82 This line of analysis is now
being extended to the international level by a number of scholars who
have focused on the nature and implications of such fundamental organiz-
ing principles of the state system as sovereignty, the balance of power, and
hegemonic domination for the conceptualization of the state and explana-
tion of state action.83
Purely schematic though these remarks are, I think all these bodies of
research would potentially contribute to a single overall problematic gener-
ated by a structurationist approach to international relations—the develop-
ment of a critical theory of the causal powers and interests of the state. An
implication of a rejection of structural monism, however, is that it will be
necessary to theorize the relationships or "articulations" between the differ-
ent structures which constitute the state. In other words, a "structure-
structure" problem emerges from the structurationist problematic. This
problem is at the core of the literature on the "articulation of modes of
production,"84 and is implicit in much of the recent "post-Marxist" debate.
Ironically, this problem is strongly reminiscent of J. D. Singer's original
81. Examples of the "productionist" critique of world-system theory include Brenner, "The
Origins of Capitalist Development," and Howe and Sica, "Political Economy, Imperialism, and
the Problem of World-System Theory." The alternative conceptualization of the structure of
the capitalist world economy (in terms of a global mode of production) has been most fully
developed by the "internationalization of capital" school of Marxist political economy; see
Christian Palloix, "The Self-Expansion of Capital on a World Scale," Review of Radical Polit-
ical Economics 9 (Summer 1977), pp. 1-28.
82. Poulantzas, State, Power, Socialism; Nicos Mouzelis, Politics in the Semi-Periphery
(New York: St. Martins, 1986); Samuel Bowles and Herbert Gintis, Democracy and Capitalism
(New York: Basic Books, 1986).
83. Bruce Andrews, "Social Rules and the State as a Social Actor," World Politics 27 (July
1975), pp. 521-40; Robert Cox, "Gramsci, Hegemony, and International Relations: An Essay in
Method," Millenium 12 (Summer 1983), pp. 162-75; Ruggie, "Continuity and Transforma-
tion"; Ashley, "The Poverty of Neo-Realism," and "Social Will and International Anarchy:
Beyond the Domestic Analogy in the Study of Global Collaboration," in Hayward Alker and
Ashley, Anarchy, Power, Community: Understanding International Cooperation (forthcom-
ing). Despite the potential usefulness of this research to the structurationist problematic, how-
ever, some of these scholars would probably reject association with that theory, especially
insofar as it is grounded in realist philosophy of science.
84. Aidan Foster-Carter, "The Modes of Production Controversy," New Left Review 107
(January-February 1978), pp. 47-77; Harold Wolpe, ed., The Articulation of Modes of Produc-
tion (London: Routledge & Kegan Paul, 1980).
discussion of the levels of analysis problem,85 although the latter would be

seen here as one instance of a more general theoretical and methodological
problem of apprehending the relationships between different structures of
whatever type (political or economic) or level of analysis (domestic or inter-
national). Although as a theoretical and methodological issue the "structure-
structure problem" points beyond structuration theory, its treatment clearly
has implications for the attempt to build satisfactory theories and explana-
tions of state action. Each structure in which the state is embedded will have
its own logic, reproduction requirements, and thus prescriptions for compe-
tent state practice. Developing a structural theory of the state and state
action, then, involves more than simply explicating the different structuring
principles which generate states; it also requires isolating and assessing the
causal role of and interrelationships among different and sometimes compet-
ing structural determinations of state action.
The need for a theory of the state in international relations is mirrored by
the need for theories of the system structures which constitute the state. In
general terms, these theories would have at least two main elements: 1) a
synchronic model of the organizing principles, logic, and reproduction re-
quirements of the structure in question, and 2) an historical account of the
genesis and reproduction of the structure. Structural theorists have gener-
ally been reluctant or unable to grasp the contingent nature of structural
genesis and reproduction, and have tended instead towards functional or
teleological readings of that process. This tendency can be corrected by the
emphasis of structuration theory on the historical specificity and contin-
gency of the structuring of social structures. Given the structurationist con-
ceptualization of social structure as only instantiated and reproduced by the
activities of social agents, an historical analysis of social structuring must
begin with the intended and unintended consequences of state action (and
the action of other agents). Although the relevant methodological tools to
such an analysis are potentially quite varied, the game-theoretic methodol-
ogy characteristic of recent work in neorealism is potentially well-adapted to
this analytical task. The analysis of iterated games and the "new insti-
tutionalism" in the study of political institutions in particular have proven
useful in generating insights into the emergence of and reproduction of social
institutions as the unintended consequences of strategic interactions,86 and
there is no a priori reason why we cannot extend the logic of such analyses
to the analysis of generative structures. We must recognize, however, that
85. J.D. Singer, "The Levels of Analysis Problem in International Relations," in Klaus
Knorr and Sidney Verba, eds., The International System: Theoretical Essays (Princeton:
Princeton University Press, 1961), pp. 77-92.
86. Michael Taylor, Anarchy and Cooperation (New York: Wiley, 1976); Robert Axelrod,
The Evolution of Cooperation (New York: Basic Books, 1984); James March and Johan Olsen,
"The New Institutionalism: Organizational Factors in Political Life," American Political Sci-
ence Review 78 (September 1984), pp. 734-48.
game-theoretic models focus attention on the technical decision problems of

given agents, and that they therefore tend to neglect the ways in which the
structure of social interactions constitute or empower those agents in the
first place. The use of game theory to develop an historical understanding of
the emergence of social structures, therefore, would have to be comple-
mented by a generative understanding of the construction of agents and
situations of strategic interaction.
These remarks on the implications of structuration theory for the scope
and content of international relations theories are obviously purely sche-
matic and are intended only to illustrate the kinds of research that might be
relevant to a structurationist approach to international relations. Indeed, I
should emphasize that structuration theory by itself cannot generate specific
theoretical claims about international relations. The theory has epistemolog-
ical implications for the form which explanations of state action should take,
and it suggests a research agenda for subsequent theorizing, but it does not
make a direct contribution to our substantive understanding of international
relations per se. This point raises the issue of the criteria by which structura-
tion theory should be evaluated by scholars of international relations. Given
that its analytical or meta-theoretical quality prevents an empirical assess-
ment of the theory, it seems to me that structuration theory should be
evaluated on pragmatic grounds, on its ability to solve problems in existing
substantive theories, to suggest new areas of theoretical and empirical in-
quiry, or to integrate different bodies of research. By this criterion, I think
structuration theory clearly impoves on its principal competitors, individ-
ualism and structuralism. It provides a framework for explaining the proper-
ties of both state agents and system structures which is denied to the
individualist and structuralist ontologies of neorealism and world-system
theory, and it defines a research agenda for international relations that orga-
nizes and subsumes under a single problematic a potentially greater variety
of extant social scientific research. The potential contribution to interna-
tional relations research is there, but we cannot assess the importance of the
contribution until theorists attempt to use a structurationist perspective to
ground and inform their theoretical and empirical research.
Conclusion
Rather than trying to summarize a long argument about the agent-structure

problem and structuration theory, I want to conclude this article by picking
up its other main thread, the implications of scientific realism for social
scientific research. Whether or not structuration theory provides a satisfac-
tory resolution to the agent-structure problem, social scientists' adoption of
a scientific realist perspective on ontology and epistemology could have
potentially revolutionary consequences for their theoretical and empirical
research. The hegemony of empiricist discourse in social science has led

social scientists into an apparent dichotomy between "science" (that is,
empiricist science) and the allegedly "un-scientific" paradigms of her-
meneutics and critical theory.87 Whatever the limits to naturalism in the
social sciences, scientific realism undermines this dichotomy by challenging
the core of the empiricist's argument, the interpretation of natural science
upon which her appropriation of the mantle of "science" rests. Scientific
realism, then, offers an alternative to the standard positions in the Positivis-
musstreit, one which enjoins social scientists to think "abductively" about
"causal mechanisms" to build their theories, instead of trying to find law-
like generalizations about observable regularities. Among the more impor-
tant consequences of such an ontological and epistemological shift is a
scientific motivation for structural theorizing in the generative or relational
sense. This methodological prescription is inherently "critical" since it re-
quires a critique and penetration of observable forms to the underlying social
structures which generate them.88 An implication of scientific realism, then,
is that "critical theory" (in a broad sense) is essential to the development of
social science, and by extension international relations, as a "science."
87. Brian Fay, Social Theory and Political Practice (London: Allen & Unwin, 1975).
88. Roy Bhaskar, "Scientific Explanation and Human Emancipation," Radical Philosophy
26 (1980), pp. 16-26; Christopher Dandeker, "Theory and Practice in Sociology: The Critical
Imperatives of Realism," Journal for the Theory of Social Behavior 13 (July 1983), pp. 195-210.

Wendt 1987

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Wendt 1987

Uploaded by

Copyright:

Available Formats

International Organization

Additional services for International

Email alerts: Click here

The agent-structure problem in international

International Organization / Volume 41 / Issue 03 / June 1987, pp 335 - 370

Link to this article: http://journals.cambridge.org/abstract_S002081830002751X

How to cite this article:

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/INO, IP address: 128.184.220.23 on 13 Mar 2015

Two theories, neorealism and world-system theory, strongly influence con-

International Organization 41, 3, Summer 1987

structured, human or organizational agents, and of their relationship to so-

1. The agent-structure problem

organizations are purposeful actors whose actions help reproduce or trans-

5. The agent-structure problem has, in various guises, recently emerged as something of a

significant. Thus, approaches to social inquiry that conceive of human be-

2. Reductionism and reification

In this section I want to: 1) compare the conceptions of "structural" theory

reductionist theories, Kenneth Waltz defines the latter as theories which

7. Waltz, Theory of International Politics, p. 18.

system structures play in neorealist explanations of state behavior. Indeed,

talist" except in terms of the individual's social relations to other agents.

14. "Reflections on Theory of International Politics: A Response to My Critics," in Robert

neo-Marxists might argue, a pure cooperation game.16 And, without a com-

approach to the agent-structure problem: they at least implicitly make one

the distribution of capabilities across preexisting states, world-system theo-

21. My generative reading of world-system theory's conceptualization of structure is charac-

neorealism's individualist approach it is able, in principle, to explain the

positions (core, semi-periphery, periphery) is functional for the reproduction

trayal of world-system theory. Indeed, in contrast to the complete neglect of

3. An alternative approach to the agent-structure problem

Structuration theory is a relational solution to the agent-structure problem

tal success of science in controlling the world.44 If mature scientific theories

make a true explanatory claim, it is necessary to identify the underlying

consists in the identification of underlying causal mechanisms rather than in

implications for the potential content of substantive theories about real-

the frequently heard complaint of mainstream social scientists, who claim

of its elements, social structures are only instantiated by the practices of

63. Bhaskar, The Possibility of Naturalism, pp. 48-49.

The importance of their internal organizational structures notwithstand-

what Giddens means by the "duality of structure," that "the structural

4. Implications for international relations theory

67. Giddens, Central Problems in Social Theory, p. 69.

of the explanation of state action, however, it is only a necessary, not a

structures and agents in turn as temporarily given in order to examine the

minate predictions about particular state behaviors. The possible ways of

talism is arguably marred by an overemphasis on exchange relationships,81

discussion of the levels of analysis problem,85 although the latter would be

game-theoretic models focus attention on the technical decision problems of

Rather than trying to summarize a long argument about the agent-structure

research. The hegemony of empiricist discourse in social science has led

You might also like